Product Manager · Seattle, WA
I build products and think in systems. My background spans backend engineering and enterprise software, giving me the technical depth to engage meaningfully in design decisions and the product instinct to translate complexity into things people actually use.
Background
I started my career as a backend engineer building systems like AML transaction monitoring, financial data pipelines, and compliance infrastructure for global banks. That foundation taught me to think in systems, not features.
I moved into product deliberately. I wanted to own the problem, not just the solution. The shift from Software Engineer to Product Owner to Product Manager was a choice I made each time I felt too comfortable.
Today I am focused on platform and technical product roles at companies building at scale. I bring engineering depth, a bias for evidence over intuition, and a genuine curiosity about how AI can make everyday work faster and less frustrating.
Career
Academics
Work
Featured
What it does
Turns a one-line idea into a structured spec covering the problem, target user, screens, and what's explicitly out of scope. It includes one editable review checkpoint, then Claude generates the prototype code directly, rendered live inside the app with Sandpack, no hosted preview dependency.
Why I built it
As a PM, turning a rough idea into something a stakeholder or engineer can respond to normally means writing a spec, then waiting on design or engineering time to see it as a real screen. I wanted to test whether AI could compress that loop into minutes, without needing to pull anyone else in first.
Result
One-line idea to clickable prototype in 1–3 minutes, reviewed by 7 people across PM, design and engineering.
📄 Case Study
Covers the three-stage build, a real PM working example, and what I learned building it, including negative prompts backfiring in image generation, cutting a stage that added no value, and why I replaced a hosted prototyping tool with Claude generating the prototype code directly, rendered in-app with Sandpack.
What it covers
An experiment comparing three AI models from three providers, Claude Sonnet 5, DeepSeek Flash, and GPT OSS 120B on Groq, across the two steps of my Idea to Prototype tool, writing the spec and writing the prototype code. Covers quality, speed, and cost per prototype, how a single setting changed DeepSeek's results, and where my own automated checks fell short.
Why I wrote it
My tool ran on a premium model that I'd never tested against alternatives. I wanted to find out whether a cheaper or faster model could do the same job, so I tested three models on quality, speed, and cost to find the most efficient choice for the tool, and to learn when routing across models is worth it.
Result
DeepSeek Flash matched Claude Sonnet 5 on quality at about 1/28th of the cost.
What it does
Compares two competing products using real user discussions from Reddit, Hacker News and DEV.to.
Why I built it
I wanted to ground competitive analysis in actual user voice rather than assumptions. Applied it to Microsoft's ecosystem to identify unmet user needs and surface product opportunities.
Result
Tested independently by 5 PMs, who found a mix of familiar and new insights.
📄 Case Study
A competitive intelligence study applying the engine to Microsoft's ecosystem — Edge AI vs Chrome AI and Teams vs Slack. Covers methodology, findings, and product recommendations grounded in real user data.
More work
What it does
Turns six product inputs into a north star, input metrics, guardrails and the vanity metrics to avoid.
Result
Returned a value-based north star on all 8 products tested.
What it does
Surfaces the assumptions, constraints and options in a high-stakes decision while leaving the final call to the user.
Result
Across 10 real decisions with 7 people, it never made the decision itself.
What it covers
An outside-in analysis of Stripe's API design, ecosystem strategy and platform metrics.
Result
Recommended guided workflow paths in test mode, organized around what developers are building, to cut time to first successful payment.
What it covers
An outside-in analysis of how Seller Central balances seller scale with buyer experience.
Result
Recommended a three-layer onboarding system that detects integration failures, alerts sellers in real time, and offers a manual fallback.
Get in touch
I'm looking for my next PM role at a company where product, engineering, and AI come together in interesting ways. If you need a PM who thinks like an engineer, has spent the past year getting hands-on with AI, and can communicate just as clearly with business stakeholders as with engineering teams, let's talk.